|Publication number||US7570726 B2|
|Application number||US 11/777,967|
|Publication date||Aug 4, 2009|
|Filing date||Jul 13, 2007|
|Priority date||Jul 14, 1999|
|Also published as||US6839393, US7466784, US7489756, US7548601, US7702057, US8428210, US20050057292, US20070258554, US20070258555, US20080013662, US20090045854, US20100146172|
|Publication number||11777967, 777967, US 7570726 B2, US 7570726B2, US-B2-7570726, US7570726 B2, US7570726B2|
|Original Assignee||Rambus Inc.|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (42), Non-Patent Citations (3), Referenced by (7), Classifications (15), Legal Events (2)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This is a continuation of U.S. application Ser. No. 10/963,828, filed Oct. 11, 2004, which is a continuation of U.S. application Ser. No. 09/353,547 filed Jul. 14, 1999, which is now U.S. Pat. No. 6,839,393, which is incorporated herein by reference.
This invention relates generally to master/slave digital systems. More particularly, this invention relates to a master/slave digital system in which the master device performs all synchronization operations, thereby obviating the need for phase alignment circuitry on slave devices.
Synchronous bus systems have been developed to support higher bandwidth digital systems. In a synchronous bus system, data packets are sent between a master device and one or more slave memory devices. The data packets travel in parallel with a system clock and maintain a precise phase to that clock. Typically, synchronous memory systems utilize a phase adjusting circuit on the master and on each one of the slave devices in order to align the on-chip receive and transmit clocks to maximize the reception timing margins both on the master and slave devices. U.S. Pat. Nos. 5,432,823 and 5,513,327, which are assigned to the assignee of the present invention and which are expressly incorporated by reference herein, describe synchronous memory systems with phase adjusting circuitry on both master and slave devices.
The primary disadvantage of using phase alignment circuitry in every slave device of a master/slave system is that it increases power dissipation. For example, if the “standby” power of phase alignment circuitry, such as a Delay Locked Loop, is 125 mW, a system with 32 slave devices on standby mode will dissipate a total of 4 W. This power dissipation is prohibitively large for emerging portable applications.
In view of the foregoing, it would be highly desirable to provide a master/slave system in which the slave devices do not require phase alignment circuitry, thereby allowing the master/slave system to operate with substantially reduced power dissipation.
A method of operating a master/slave system includes the step of identifying a master receive data phase value to coordinate the transfer of data from a slave device without phase alignment circuitry to a master device with a universal phase aligner. Data is transferred from the slave device to the master device in accordance with the master receive data phase value. The master device characterizes a master transmit data phase value to coordinate the transfer of data from the master device to the slave device. Subsequently, the master device routes data to the slave device in accordance with the master transmit data phase value.
A master/slave system includes a set of slave devices, with each slave device having a clock circuit without phase alignment circuitry. A master device with a universal phase alignment circuit includes a phase value register bank storing a set of phase values for the set of slave devices. The master device utilizes a selected phase value of the set of phase values to alter a system clock signal in accordance with the selected phase value so as to establish synchronous communication between the master device and a selected slave device of the set of slave devices.
The invention provides a master/slave system in which the slave devices do not require phase alignment circuitry. Thus, the master/slave system operates with substantially reduced power dissipation. The apparatus of the invention can be operated in a high throughput mode wherein each slave device is assigned a first phase delay value for data transmission and a second phase delay value for data reception. In a simpler, lower throughput mode, individual phase delay values are not stored. Instead, the master designates different slave devices to process different data edges, depending upon the phase information the master device receives during a calibration operation.
For a better understanding of the invention, reference should be made to the following detailed description taken in conjunction with the accompanying drawings, in which:
Like reference numerals refer to corresponding parts throughout the drawings.
In the embodiment of
The universal phase aligner 30 also produces a master receive data clock (MRDCLK) signal, which is used to control a latch 36, which may be a flip-flop. The latch 36 latches data from the data bus 42. The data bus 42 carries data to and from the master device 22 (i.e., the data bus 42 is bi-directional). A latch is provided for each line of the data bus 42.
As discussed below, the MRDCLK signal is a phase adjusted system clock signal for a selected slave device 24. The MRDCLK signal is synchronized to data received from the selected slave device of the master/slave system 20. The MRDCLK signal accounts for the time between when the system clock is used to launch data from a slave device 24 and the time that the data is received at the master device. By accounting for this time difference, expressed as a phase difference between the system clock and the MRDCLK, the slave device 24 and the master device 22 can exchange data in phase.
The universal phase aligner 30 also produces a master transmit data clock (MTDCLK) signal, which controls the timing of an output buffer 38. The output buffer 38 drives data from the logic and control circuitry 32 onto the data bus 42. An output buffer 38 is provided for each line of the data bus 42.
As discussed below, the MTDCLK signal is a phase adjusted system clock signal for a selected slave device 24. The MTDCLK signal is synchronized to transmit data, such that the selected slave device of the master/slave system 20 receives the data in phase with the system clock. The MTDCLK signal accounts for the time between launching data from the master 22 to the time that it is received at the slave 24. This time difference, expressed as a phase difference between the system clock and the MTDCLK, allows the master device 22 to launch data in a manner such that it is received at a selected slave device 24 in phase with the system clock.
Each slave device 24 of the master/slave system 20 includes a clock circuit 50 and a set of logic and control circuitry 52. The clock circuit 50 does not include any type of phase alignment circuitry. Instead, it is merely implemented with an amplifier, a buffer, and duty cycle correction circuitry, if necessary. The amplifier and buffer may be used to amplify and buffer the system clock (SYSCLK) received from line 33. The amplified and buffered system clock (SCLKO) is used to gate a request bus flip-flop 54, a data bus input flip-flop 56, and a data bus output buffer 58. The request bus flip-flop 54 is attached to the request bus 40, while the data bus input flip-flop 56 and the data bus output buffer 58 are connected to the data bus 42. Flip-flops 56 and output buffers 58 are provided for each line of the request bus 40 and data bus 42. The implementation of the logic and control circuitry 52 is inapposite to the invention, as a variety of logic and control functions may be implemented in accordance with the synchronization scheme of the invention.
The operations performed by the components of
The universal phase aligner 30 may include a standard system clock source 60. Alternately, an external system clock may be used. The system clock signal (SYSCLK) is applied to line 33 of
The output of the phase rotator 62 is an adjusted system clock signal, which may be the previously described master request clock (MRQCLK), master receive data clock (MRDCLK), or the master transmit data clock (MTDCLK). The adjusted system clock signal is applied to a flip-flop 64, which also receives a clock signal from a selected slave device. In particular, a signal is received at latch 36 of
A counter 68 is used to quantify the phase difference between the adjusted clock signal and the clock signal received from the selected slave device. For example, the counter 68 may be implemented to count digital values read from the flip-flop 64. In such an embodiment, the flip-flop 64 may be timed by a digital high adjusted clock signal. A digital high output from the flip-flop 64 indicates that the clock signal from the slave device is at a digital high value at the same time as the adjusted clock signal. Thus, the signal from the slave device has arrived at the flip-flop 64. On the other hand, a low output from the flip-flop 64 indicates that the clock signal from the slave device has not been received at the flip-flop 64. In such a case, the counter is incremented to adjust for the phase error. When phase alignment is achieved, the counter 68 stores a counter value indicating the phase difference between the adjusted clock signal and the clock signal received from the selected slave device.
At this time, a controller 70 reads the counter value in the counter 68. The phase difference value read from the counter is then stored in a phase value register bank 74. In particular, the phase difference value is stored in a register corresponding to the selected slave device. Recall that the stored value specifies a phase value adjustment to the system clock that will match the clock signal received from the selected slave device to the adjusted clock signal. Since the signals are matched, each signal transition will occur at the same time. This may result in metastability problems as the signals are latched into different devices. To avoid this problem, an offset adder 72 is used to add an offset value to the phase value received from the counter 68. The offset improves data reception timing margins. By way of example, the offset may be 90 degrees.
Later when the master 22 communicates with a selected slave device 24, the phase value for the selected slave device is retrieved by the controller 70 from the phase value register bank 74. The controller 70 then loads the value into the counter 68 and disables the counter. Thus, the same phase value is applied to the phase rotator 62 with every clock cycle. This results in a phase rotation of the system clock (SYSCLK) to correspond to the specified phase value.
Each of the components of
As previously indicated, the master/slave system of
The master transmit data calibration operation is performed to identify the phase value to be used to transmit data to a selected slave device. The phase value secured during the master transmit data calibration operation is used during a master transmit data operation. That is, the phase value is used to adjust the system clock so that an output buffer of the master device 22, operating in response to the adjusted system clock, transmit data that is in phase with the system clock as received at the selected slave device.
The master data receive calibration operation 80 is disclosed in connection with
The next processing step of
A decision is then made to determine whether all of the slave devices have been calibrated (step 92). If not, a calibration operation for the next slave device in the system is initiated with step 82. Once all of the slave devices in the system have been calibrated to obtain a data receive phase value, the calibration operation 80 is completed.
The master transmit data calibration operation 110 of the invention is disclosed in connection with
The master device 22 then aligns the master request clock (MRQCLK) until it corresponds to the re-routed request bus data at the slave device 24. Observe that in this case, the phase difference created by the distance on the data bus 42 from the selected slave device 24 to the master device 22 is accounted for by the data receive phase value for the selected slave device, which was obtained during the master data receive calibration. Thus, the present calibration operation measures the phase difference created by data traveling from the master device 22 to the selected slave device 24. That is, the present calibration operation measures the phase difference created by data traveling on the request bus 40 between the master device 22 and the selected slave device 24.
In sum, the master device 22 aligns the MRQCLK signal with the re-routed request bus data to form a master transmit phase aligned MRQCLK signal for the selected slave device (step 114). As previously discussed in connection with
After alignment is achieved, the master has quantified the phase value between the system clock and the master transmit phase aligned MRQCLK signal (step 116). The master then applies a margin offset to the phase value (step 118). As previously discussed, this is performed by the offset adder 72 to improve timing margins. The master then stores the offset adjusted transmit data phase value for the selected slave device (step 120). If all of the slave devices have not been calibrated in this manner (step 122), then the processing of
The master transmit data operation 130 is executed in accordance with the method of
In sum, after the master data receive calibration operation of
Attention presently turns to the master device's transfer of communication between different slave devices 24. There is a latency associated with any such transfer of control. The factors that affect hand-over or switching latency include: (1) switching the phase of the master clock to the appropriate phase for the active slave and (2) informing a slave that it is currently active or inactive. The operation of switching the phase of the master clock to the appropriate phase for the active slave can be limited to below 1-2 cycles by pipelining techniques (e.g., switching the phase of the MRQCLK before switching the MRDCLK).
There are a number of techniques that may be used to inform a slave that it is currently active or inactive. For example, a signal on the side band bus 46 may be used. While this is an acceptable solution in some systems (e.g., graphics memory subsystems), this solution might not be tolerable in systems where the master addresses different slaves frequently (e.g., main memory systems with data fragmentation).
Another technique for advising a slave device of its status is to use the request bus 40. In this embodiment, an additional “broadcast” bit is needed by the request bus 40. To accomplish an “active hand-over”, the master 22 transmits a “broadcast” packet on the request bus 40. This packet is transmitted with a pre-determined phase of the MRQCLK and is valid for a full clock cycle (i.e., two bit time periods). All the inactive slaves 24 continuously sample the data on the request bus 40 with both of their negative and positive clock edges. However, they always disregard the data sampled with one of the two edges according to their local phase. In other words, depending upon the position of the slave device in the physical system, it will sample one of the two edges of the data. Thus, the slave device is effectively operating at half-frequency during this state. The decision as to which datum to process and which to disregard is made by the master and is communicated to the slave during the calibration phase. Since the broadcast bit is valid for a full clock cycle, one of the two slave clock edges is appropriate for receiving the broadcast bit and the data on the request bus 40 with adequate timing margin.
When the broadcast bit is a logical ONE, the master 22 broadcasts data on the request bus 40 specifying which slave is being invoked. Each slave 24 processes this broadcast data to determine if the broadcast data matches its specified identification value. If the comparison results in a match, the slave goes to the active mode. In the active mode, data on the request bus 40 and data bus 42 are processed at full speed (i.e., on both clock edges).
An additional consideration for the active “hand-over” is the amount of time that the slaves might have to wait before they start transmitting data on the bus so that their data will not collide with the data transmitted by the previously active slave (or reflections of that data). This amount of time depends on the electrical length of the bus.
The system of the invention can operate in a low power mode that is analogous to the active hand-over mode just discussed. In the low power mode, the phase of the master clock is pre-determined (e.g., it is locked to the system clock). Thus, the master does not change the phase of its clocks according to which slave is active. Unlike the previous high bandwidth embodiment in which two bits of data are transmitted every clock cycle, in this embodiment only one datum is transmitted per clock cycle.
During calibration for this mode, the master informs each slave which of its two clock edges to use when it receives and transmits data to and from the request bus 40 and data bus 42. The master chooses the slave transmit clock edge that produces the maximum timing margin on the master's receivers. The maximum timing margin is identified during calibration. Since the phase of the normal receive clock of the master is pre-determined and the position of the slave's clock edges can be found in the calibration period, the identification of the maximum timing margin is simply based on a digital comparison at the master device 22. Similarly, the master device 22 chooses the positive or negative edge of the slave's receive clock that results in the maximum timing margin on the slave receivers. Both the master transmit clock phase and the slave edge positions can be determined during the calibration period.
The highest peak bandwidth achievable by this system is one half of the synchronous bus system that transmits two bits of data per clock cycle. Nevertheless, its lower power consumption and simpler control scheme make this embodiment appealing for selected applications.
Those skilled in the art will recognize many alternate embodiments for the disclosed invention. For example, the invention was described in connection with a calibration technique that relied upon a clock signal. The calibration operation may also be performed with a random signal, as long as the universal phase aligner 30 can identify periodic signal edges associated with the random signal.
The foregoing description, for purposes of explanation, used specific nomenclature to provide a thorough understanding of the invention. However, it will be apparent to one skilled in the art that the specific details are not required in order to practice the invention. In other instances, well known circuits and devices are shown in block diagram form in order to avoid unnecessary distraction from the underlying invention. Thus, the foregoing descriptions of specific embodiments of the present invention are presented for purposes of illustration and description. They are not intended to be exhaustive or to limit the invention to the precise forms disclosed, obviously many modifications and variations are possible in view of the above teachings. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. It is intended that the scope of the invention be defined by the following claims and their equivalents.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4481625||Oct 21, 1981||Nov 6, 1984||Elxsi||High speed data bus system|
|US5097489||May 19, 1989||Mar 17, 1992||Tucci Patrick A||Method for incorporating window strobe in a data synchronizer|
|US5432823||Jan 7, 1994||Jul 11, 1995||Rambus, Inc.||Method and circuitry for minimizing clock-data skew in a bus system|
|US5485490||Nov 29, 1994||Jan 16, 1996||Rambus, Inc.||Method and circuitry for clock synchronization|
|US5513327||Mar 31, 1994||Apr 30, 1996||Rambus, Inc.||Integrated circuit I/O using a high performance bus interface|
|US5554945||Feb 15, 1994||Sep 10, 1996||Rambus, Inc.||Voltage controlled phase shifter with unlimited range|
|US5577236||Dec 30, 1994||Nov 19, 1996||International Business Machines Corporation||Memory controller for reading data from synchronous RAM|
|US5579352||Apr 6, 1994||Nov 26, 1996||National Semiconductor Corporation||Simplified window de-skewing in a serial data receiver|
|US5614855||Aug 21, 1995||Mar 25, 1997||Rambus, Inc.||Delay-locked loop|
|US5615358||Jun 7, 1995||Mar 25, 1997||Texas Instruments Incorporated||Time skewing arrangement for operating memory in synchronism with a data processor|
|US5692165||Sep 12, 1995||Nov 25, 1997||Micron Electronics Inc.||Memory controller with low skew control signal|
|US5757786||Dec 27, 1994||May 26, 1998||Hyundai Electronics Industries Co., Ltd.||Time synchronization apparatus and a method thereof using a global positioning system of a satellite|
|US5852640||Jun 26, 1995||Dec 22, 1998||Kliza; Phillip S.||Clock distribution apparatus with current sensed skew cancelling|
|US5864592 *||Nov 3, 1992||Jan 26, 1999||Pairgain Technologies, Inc.||Timing recovery system for digital subscriber line transceivers|
|US5889824||May 29, 1997||Mar 30, 1999||Nec Corporation||Intermittent receiving apparatus capable of reducing current consumption|
|US5896347||Sep 8, 1997||Apr 20, 1999||Fujitsu Limited||Semiconductor memory system using a clock-synchronous semiconductor device and semiconductor memory device for use in the same|
|US5918040||Jul 25, 1995||Jun 29, 1999||Cabletron Systems, Inc.||Method for maintaining time synchronization between two processors in a network interface|
|US5933623||Oct 25, 1996||Aug 3, 1999||Hitachi, Ltd.||Synchronous data transfer system|
|US6016282||May 28, 1998||Jan 18, 2000||Micron Technology, Inc.||Clock vernier adjustment|
|US6043694||Jun 24, 1998||Mar 28, 2000||Siemens Aktiengesellschaft||Lock arrangement for a calibrated DLL in DDR SDRAM applications|
|US6125419 *||Jun 13, 1997||Sep 26, 2000||Hitachi, Ltd.||Bus system, printed circuit board, signal transmission line, series circuit and memory module|
|US6131149 *||Mar 23, 1999||Oct 10, 2000||Oak Technology, Inc.||Apparatus and method for reading data from synchronous memory with skewed clock pulses|
|US6232806||Oct 21, 1998||May 15, 2001||International Business Machines Corporation||Multiple-mode clock distribution apparatus and method with adaptive skew compensation|
|US6236623||Oct 16, 1998||May 22, 2001||Moore Industries||System and method for synchronizing clocks in a plurality of devices across a communication channel|
|US6292903||Jun 29, 1998||Sep 18, 2001||International Business Machines Corporation||Smart memory interface|
|US6426984||May 7, 1999||Jul 30, 2002||Rambus Incorporated||Apparatus and method for reducing clock signal phase skew in a master-slave system with multiple latent clock cycles|
|US6510503||Oct 30, 1998||Jan 21, 2003||Mosaid Technologies Incorporated||High bandwidth memory interface|
|US6513103||Oct 10, 1997||Jan 28, 2003||Rambus Inc.||Method and apparatus for adjusting the performance of a synchronous memory system|
|US6553472||Jan 12, 2001||Apr 22, 2003||Sun Microsystems, Inc.||Method for programming clock delays, command delays, read command parameter delays, and write command parameter delays of a memory controller in a high performance microprocessor|
|US6611905||Jun 29, 2000||Aug 26, 2003||International Business Machines Corporation||Memory interface with programable clock to output time based on wide range of receiver loads|
|US6804764||Jan 22, 2002||Oct 12, 2004||Mircron Technology, Inc.||Write clock and data window tuning based on rank select|
|US6807614||Mar 14, 2002||Oct 19, 2004||Shine C. Chung||Method and apparatus for using smart memories in computing|
|US6928571||Sep 15, 2000||Aug 9, 2005||Intel Corporation||Digital system of adjusting delays on circuit boards|
|US6970988||Jul 19, 2002||Nov 29, 2005||Chung Shine C||Algorithm mapping, specialized instructions and architecture features for smart memory computing|
|US7076745||Feb 18, 2004||Jul 11, 2006||Oki Electric Industry Co., Ltd.||Semiconductor integrated circuit device|
|US7100066||Dec 27, 2002||Aug 29, 2006||Lg Electronics Inc.||Clock distribution device and method in compact PCI based multi-processing system|
|US7224595||Jul 30, 2004||May 29, 2007||International Business Machines Corporation||276-Pin buffered memory module with enhanced fault tolerance|
|US20010013802||Jul 7, 1999||Aug 16, 2001||Ghene Faulcon||System and process for high speed interface clock skew correction|
|US20010047450||Oct 30, 1998||Nov 29, 2001||Peter Gillingham||High bandwidth memory interface|
|JPH0387907A||Title not available|
|JPH04117709A||Title not available|
|JPS63276935A||Title not available|
|1||Chang, K.Y., "Design of a CMOS Asymmetric Serial Link," Ph.D. Dissertation, Stanford University, Aug. 1999.|
|2||Chang, K.Y., et al., "A 2Gb/s/pin CMOS Asymmetric Serial Link," Symposium on VLSI Circuits Digest of Technical Papers, 1998.|
|3||Sidiropoulos, S., "A Semidigital Dual Delay-Locked Loop," IEEE J. Solid State Circuits, vol. 32, No. 11, Nov. 1997.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7683692 *||Sep 21, 2006||Mar 23, 2010||St-Ericsson Sa||Bus circuit|
|US7872517 *||Mar 22, 2010||Jan 18, 2011||St-Ericsson Sa||Bus circuit|
|US8069379||Nov 30, 2009||Nov 29, 2011||Rambus Inc.||Memory system with point-to-point request interconnect|
|US20090085631 *||Sep 21, 2006||Apr 2, 2009||Nxp B.V.||Bus circuit|
|US20100077267 *||Nov 30, 2009||Mar 25, 2010||Rambus Inc.||Memory System with Point-to-Point Request Interconnect|
|US20100189287 *||Mar 22, 2010||Jul 29, 2010||Xavier Lambrecht||Bus circuit|
|US20100211748 *||Apr 11, 2008||Aug 19, 2010||Rambus Inc.||Memory System With Point-to-Point Request Interconnect|
|International Classification||H03K5/01, H03L7/081, G06F13/42, H04L25/00|
|Cooperative Classification||H03L7/0814, G06F1/10, G06F13/4217, G06F13/4243, Y02B60/1235, Y02B60/1228|
|European Classification||G06F13/42C3S, H03L7/081A1, G06F1/10, G06F13/42C1S|
|Dec 31, 2012||FPAY||Fee payment|
Year of fee payment: 4
|Feb 6, 2017||FPAY||Fee payment|
Year of fee payment: 8